OpenDroneMap (ODM) — What It Does

ODM is a free, open-source command-line tool that turns ordinary 2D photos taken from a drone (or balloon/kite) into useful 3D and map products. You give it a folder of images; it gives you back:

• Classified point clouds — a dense cloud of 3D points, each tagged with a category (e.g., ground, vegetation, building).
• 3D textured models — a solid 3D surface with the original photos "painted" onto it.
• Georeferenced orthorectified imagery — a flat, distortion-corrected map image where every pixel has real-world coordinates.
• Georeferenced digital elevation models — a grid of ground heights, also tied to real-world coordinates.

It runs on Windows, Mac, and Linux, and is designed for people comfortable with a terminal — power users, scripts, and integrations with other software. If you prefer a clickable interface, the project points you to a separate companion app called OpenDroneMap Desktop.

How You Use It

The simplest path is Docker. After installing Docker, you pull the ODM image, put your photos (JPEG, TIFF, or DNG) into a folder named images, and run one command that mounts your dataset folder into the container. You can append extra options — for example, --dsm to generate a digital surface model, or --orthophoto-resolution 2 to double the map resolution.

A native Windows installer also exists; after installing, you type run C:\path\to\project in the ODM console.

What Comes Out

After processing, the output files are organized into subfolders. Key files include:

• odm_mesh.ply — a raw 3D mesh
• odm_textured_model.obj / odm_textured_model_geo.obj — textured mesh, with and without geographic coordinates
• odm_georeferenced_model.laz — a compressed point cloud (LAZ format)
• odm_orthophoto.tif — the orthophoto as a GeoTIFF

The README suggests free viewers: QGIS for GeoTIFFs, CloudCompare for LAZ, and MeshLab for OBJ/PLY. It warns that ordinary photo editors like Photoshop won't open the GeoTIFFs correctly.

Extra Capabilities

• GPU acceleration — ODM can run SIFT feature extraction on an NVIDIA GPU (CUDA-based, GTX 9xx or newer) using a special :gpu Docker image and the --gpus all flag. The README notes this is roughly 2× faster than CPU on a typical laptop.
• Video input — since version 3.0.4, you can drop video files (.mp4, .mov, .lrv, .ts) into the images folder and ODM will extract frames automatically. Matching .srt subtitle files with GPS data are also supported.
• Network API — via a separate project called NodeODM, ODM can be exposed as a web service.

For Developers

The project is built with a tool called pixi. After installing it, you can build, test, and run ODM from source with commands like pixi run build, pixi run test, and pixi run odm. It relies on several other open-source libraries — notably OpenSfM, OpenMVS, PDAL, Entwine, GDAL, and GRASS GIS — and welcomes contributions via pull requests.

Community & Docs

Full tutorials live at docs.opendronemap.org, and there's an active community forum for questions and discussion.